Tools & Platforms
Rental scooter companies in downtown Knoxville want to use AI to correct parking
KNOXVILLE, Tenn. (WVLT) – If you’ve had to dodge a scooter left on the sidewalk in downtown Knoxville— you’re not alone. The rental companies who run the scooters are bringing in new technology to fix that.
Now, there’s AI coming into the Bird and Lime apps to keep the scooters where they need to be.
“When it’s right here on the ground and any person can trip and fall no one really cares where they put them down,” Candace Hearron said.
City leaders in a workshop meeting this week announced the city’s scooter partners, Lime and Bird, will have the system review pictures of the scooters when riders are done.
“So if somebody tried to end their scooter ride and leave the scooter right there, it would actually won’t let them. It will say you can’t park your scooter on the sidewalk. Try again,” Chad Weth, deputy chief of operations with the City of Knoxville, said. “You have to go park this in a different location so there is becoming more enforcement with the programs through the app and through the user.”
There’s still appreciation for the scooters.
“It’s convenient for a lot of people, it makes less car traffic,” said Hearron.
But riders using the scooters safely is another concern.
“Just going down here on Gay Street, you’re going down in your car as you’re suppose to,” she said. “But then you have people whip through and jump in front of you which can cause accidents behind you.”
As for making sure people use the scooters safely — the city and scooter companies are still looking into how to address that.
Lime’s app has the AI system running now, while Bird’s app will have the system running by the fall.
Copyright 2025 WVLT. All rights reserved.
indystar.com wants to ensure the best experience for all of our readers, so we built our site to take advantage of the latest technology, making it faster and easier to use.
Unfortunately, your browser is not supported. Please download one of these browsers for the best experience on indystar.com
Opinion | Governing AI The Kathmandu Post
Source link
Switzerland’s EPFL launched the open multilingual Apertus LLM that comes in two sizes, 8B and 70B, and targets massively multilingual coverage, with over 1800 languages supported. The models were trained on 15 trillion tokens across more than 1,000 languages; 40% of the data is non-English. Apertus is fully open, sharing weights on HuggingFace as well as data, training script details and a Technical Report.
Nous Research released Hermes 4 14B, a hybrid-reasoning LLM positioned as the compact sibling to its 70B and 405B variants. The 14B model can run locally (available on HuggingFace) while supporting dual “think/non-think” modes and function-calling support. Nous Research published a technical report sharing training and benchmark details.
Nous also announced the Husky Hold’em Bench, a poker-themed benchmark designed to test long-horizon reasoning and strategy under uncertainty. The suite provides a consistent evaluation scaffold for agentic AI models.
Tencent announced the open-source release of Hunyuan-MT-7B, a multilingual translation model that supports 33 languages, boasts lightning-fast high-performance inference for translation, and can be deployed on edge devices. They also released an ensemble “Chimera” variant:
But that’s not all. We’re also open-sourcing Hunyuan-MT-Chimera-7B, the industry’s first open-source integrated translation model. It intelligently refines translations from multiple models to deliver more accurate and professional results for specialized use cases.
Both Hunyuan-MT-7B and Hunyuan-MT-Chimera-7B are available on HuggingFace.
Alibaba’s Tongyi Lab released WebWatcher, an open research agent that can browse and complete web tasks end-to-end. The HuggingFace demo and WebWatcher code and technical paper are provided for reproducibility. It is positioned as a reference for building web-capable agents.
Google introduced EmbeddingGemma, a small 308M-parameter multilingual embedding model for on-device RAG, search and similarity workloads. EmbeddingGemma documentation emphasizes small memory footprint with strong MMTEB performance relative to its size.
Google launched Androidify, a playful generative tool for custom Android avatars. Powered by Gemini 2.5 Flash and Imagen, Androidify lets users generate Android-style characters, remix them into sticker packs, and share to social apps and Messages. Google positions it as an “AI sandbox” for creative expression.
GitHub shipped Actions for AI Labeler and AI Content Moderator, powered by GitHub Models. The Actions let repos auto-apply labels or flag content in workflows, reducing toil in triage and moderation. GitHub positions them as drop-in CI/CD components with configurable prompts and thresholds.
Mistral added enterprise connectors and a Memories capability to Le Chat to integrate data sources and persist knowledge for better context. The update expands Le Chat beyond chat to more agentic workflows. Mistral’s post outlines new features and supported integrations.
OpenAI launched Projects in ChatGPT, a feature similar to Claude Workspaces that adds structured workspaces, larger file uploads, and more granular memory controls. They also added a “Branching” feature to ChatGPT to fork a conversation into alternative directions without losing the original thread.
Apple’s FastVLM family (0.5B/1.5B/7B) is now available on Hugging Face, providing faster vision encoding via the FastViTHD encoder. The FastVLM model was presented in the FastVLM paper published last December.
Luxury membership platform Vivrelle launched Ella, an AI personal styling tool, in partnership with fashion retailers Revolve and FWRD. Ella offers personalized outfit and styling recommendations across all platforms, covering rental, retail, and pre-owned options.
Google’s latest video-generation model, Veo 3, is coming to Google Photos. This new model allows U.S. users to turn still images into higher-quality four-second video clips (without audio) via the mobile app’s Create tab.
OpenAI published a research explainer on why language models hallucinate. The team argues current training evaluation incentives reward “confident guessing” over calibrated uncertainty; they advocate benchmarks that penalize wrong answers more than abstentions. The paper ties hallucinations to the statistical nature of next-token prediction and calls for uncertainty-aware evaluation.
Google DeepMind revealed how Deep Loop Shaping improves gravitational wave observations, helping scientists understand the universe better. In experiments at LIGO Livingston, Deep Loop Shaping reinforcement-learning method reduced noise in a critical feedback loop by 30–100 times, enabling detection of many more gravitational-wave events. DeepMind says the technique could be applied to vibration and noise suppression in aerospace and robotics.
Anthropic raised $13B at a $183B post-money valuation, citing $5B run-rate revenue and 300k business customers. The company says Claude Code passed $500M run-rate within months of full launch; funds will support enterprise demand, safety research, and global expansion.
Anthropic agreed to a landmark $1.5B settlement with authors over training data and opt-out rights. The deal, awaiting court approval, creates a claims process, an opt-out mechanism for authors, and a remuneration program for future training use. Over 500,000 writers are eligible for payouts. This record payout is for illegally pirating books to train its Claude AI model, not for the act of training on copyrighted material itself. It’s the most sweeping U.S. copyright settlement related to AI training to date.
Microsoft announced an agreement with the Federal Government to scale AI adoption, including Microsoft Copilot and Azure OpenAI. Microsoft AI services have achieved security and compliance authorizations, including FedRAMP High and DoD provisional authorization for Microsoft 365 Copilot, to support rapid AI rollout across Federal agencies.
Broadcom’s quarterly shows accelerating AI chip revenue and a blockbuster new customer order that could be OpenAI. The company said a new customer placed over $10 billion in AI infrastructure orders, fueling speculation about OpenAI ties. OpenAI is reportedly partnering with Broadcom to launch its first in-house AI chip in 2026, as it aims to diversify compute away from off-the-shelf GPUs.
OpenAI has had a busy week of aquisitions, deals, reorganization, and launches:
CoreWeave agreed to acquire OpenPipe, a startup known for reinforcement-learning tooling for agents. The companies announced a definitive agreement. This helps build out CoreWeave’s technology stack to support AI use cases.
New York Fed survey finds AI adoption up sharply, but job impacts limited, for now. Roughly 40% of services firms and 26% of manufacturers in the New York region report using AI, with retraining more common than layoffs; firms expect both hiring and some future reductions as AI adoption deepens.
Warner Bros. is suing AI startup Midjourney for copyright infringement, alleging it allows users to generate copyrighted characters like Superman and Batman without permission. The lawsuit claims Midjourney knowingly lifted content restrictions, profiting from “piracy and copyright infringement.”
California and Delaware Attorneys General are investigating OpenAI following reports of tragic deaths potentially linked to ChatGPT interactions, demanding improved safety measures for young users. In related news, both OpenAI’s ChatGPT and Google’s Gemini AI products have raised significant child safety concerns, with Common Sense Media rating Gemini as “High Risk” due to the potential for inappropriate content and unsafe advice.
Concerns about inappropriate interactions and lack of safety features in AI for children could drive further regulatory action. The FTC is making inquiries into how AI chatbots affect children’s mental health, according to WSJ reporting. The agency may seek documents from OpenAI, Meta and Character.AI.
Tesla shareholders will vote on investing in Elon Musk’s xAI to bolster its AI, robotics, and energy ambitions. Proposed by a shareholder, this move aims to secure Tesla’s stake in advanced AI capabilities and drive value.
AI agent startup Sierra, co-founded by Bret Taylor, raised $350 million in funding, valuing the company at $10 billion. Sierra helps enterprises build customer service AI agents.
Harish Abbott’s Augment, featuring its AI assistant “Augie” for logistics, secured an $85 million Series A round led by Redpoint. Augie automates repetitive tasks like gathering bids, tracking packages, and managing invoices for freight companies.
Isotopes AI came out of stealth on Thursday with a $20 million seed round. Its AI agent, Aidnn, enables business managers to query complex data in natural language, drafting planning documents from various sources like ERP and CRM.
OpenAI lays out its case for AI’s role in expanding economic opportunity. This company essay by Applications CEO Fidji Simo is a strategic positioning piece ahead of the launch of their Jobs platform, to explain push for AI literacy through certifications and their Jobs Platforms. He argues that AI broadens access to income-generating tools, but that AI’s disruption requires reskilling at scale.
The premise for AI reskilling is simple:
Studies show(opens in a new window) that AI-savvy workers are more valuable, more productive, and are paid more than workers without AI skills.
Yes, AI literacy is vital to being successful in the AI era, and soon most American workers will interface with AI at work. The AI technology revolution is so fast that even OpenAI’s stated goal of 10 million certifications by 2030 seems insufficient.
Our motivation for writing AI Changes Everything is similar; we want you to help you stay on top of AI. If there is a topic you want me to dive further into for your own AI up-skilling, leave a comment and I’ll dig into it.
-
Business1 week agoThe Guardian view on Trump and the Fed: independence is no substitute for accountability | Editorial
-
Tools & Platforms4 weeks ago
Building Trust in Military AI Starts with Opening the Black Box – War on the Rocks
-
Ethics & Policy1 month agoSDAIA Supports Saudi Arabia’s Leadership in Shaping Global AI Ethics, Policy, and Research – وكالة الأنباء السعودية
-
Events & Conferences4 months agoJourney to 1000 models: Scaling Instagram’s recommendation system
-
Jobs & Careers2 months agoMumbai-based Perplexity Alternative Has 60k+ Users Without Funding
-
Education2 months agoVEX Robotics launches AI-powered classroom robotics system
-
Podcasts & Talks2 months agoHappy 4th of July! 🎆 Made with Veo 3 in Gemini
-
Funding & Business2 months agoKayak and Expedia race to build AI travel agents that turn social posts into itineraries
-
Education2 months agoMacron says UK and France have duty to tackle illegal migration ‘with humanity, solidarity and firmness’ – UK politics live | Politics
-
Podcasts & Talks2 months agoOpenAI 🤝 @teamganassi